Seeing water split a drug–polymer dispersion: two-dye confocal microscopy of miconazole–PVPVA films

Background and motivation

An amorphous solid dispersion (ASD) dissolves a poorly soluble drug, in amorphous form, in a polymer matrix. Its advantages rest on one premise: drug and polymer are mixed at the molecular level. If they demix into drug-rich and drug-lean domains, the polymer can no longer hold back crystallization in the drug-rich regions, and both stability and dissolution suffer (Alonzo 2010; Ilevbare 2012; Konno and Taylor 2006). C1 When the demixing leaves both components amorphous, it is called amorphous–amorphous phase separation (AAPS). AAPS has been found to come before crystallization, because the drug-rich phase carries less polymer (Rumondor 2009b), and crystallization in turn removes the solubility advantage of the amorphous form (Yu 2001). C2

Water is the usual trigger. AAPS has been reported during manufacturing (Qian 2010), from moisture absorbed during storage (Marsac 2010; Rumondor 2009b) and from bulk water during dissolution (Purohit and Taylor 2015b), where separation can happen within seconds of hydration. C3·C33·R26 Water in the processing solvent can demix drug and polymer while the ASD is being formed, because water is an antisolvent for the drug (Li 2017), much as an antisolvent for one polymer separates polymer blends (Hellmann 1991; Pouchlý and Patterson 1976; Robard 1977). C32·R28 Some systems are very sensitive to water on storage and others are not (Rumondor 2009c, 2010, 2011; Vasanthavada 2005). C33 The authors point out that these observations come from different drug–polymer systems, so no study had followed water entering at different stages in the same system. C34

A second problem is detection. DSC can miss AAPS because domains may remix during the heating scan (Qian 2010) and because it cannot resolve domains below about 30 nm (Newman 2008); X-ray diffraction is suited to long-range order; atomic force microscopy (AFM) and electron microscopy resolve nanometre features but cannot tell drug from polymer; IR and Raman spectroscopy report drug–polymer interactions, but their interpretation is not simple. C4·R27 AFM combined with nanoscale IR (AFM-IR) and with Lorentz contact resonance (LCR) had been used for telaprevir dispersions (Li and Taylor 2016). C4 Fluorescence methods were emerging (Purohit and Taylor 2015a, b; Tian 2016), but most studies stained only the drug-rich phase with a hydrophobic dye. In cell biology several dyes, each for a different structure, are routine, and the authors' group had just shown that the same can work for drug–polymer systems, because the polymer is usually the more hydrophilic component (Purohit 2017). C5

The model system was miconazole with poly(vinylpyrrolidone-co-vinyl acetate) (PVPVA, Kollidon VA 64). Its miscibility depends on how the dispersion is made and stressed: spray-dried dispersions above 20% drug were phase separated (Singh 2014), melt-extruded ones mixed better and showed a single glass transition (Singh 2015), and compression could remix separated material. C6·R24 The paper therefore had two aims: to test confocal fluorescence microscopy (CFM) with a drug-specific and a polymer-specific dye as a miscibility tool, and to use it to follow how the microstructure of miconazole–PVPVA films develops when water enters during preparation, storage at high relative humidity (RH) and dissolution. C7 The working hypothesis was that a hydrophobic dye would partition into the drug-rich phase and a hydrophilic dye into the polymer-rich phase, giving both phases contrast in one image. C7

Background concepts

Amorphous–amorphous phase separation. In an AAPS the single amorphous mixture becomes two amorphous phases of different composition: here a miconazole-rich and a PVPVA-rich phase. Neither phase is crystalline, so X-ray diffraction sees no change, and DSC sees it only if the domains are large enough and do not remix during heating (two glass transitions instead of one). C4·C23 The arrangement of the two phases matters for this paper. One phase can form discrete droplets in a continuous matrix of the other, or both phases can form connected networks at the same time, which is called a bicontinuous structure. C28

Nucleation and growth. When a mixture separates by nucleation and growth, small droplets of the new phase appear and grow inside the original matrix, so the result is discrete domains in a continuous phase. The authors use this picture to read the discrete drug-rich droplets seen after humid casting. C40

Mobility versus thermodynamics. Absorbed water can raise the molecular mobility of an amorphous solid, which is reflected in its glass transition temperature (Tg) (Hancock and Zografi 1994). Water can also change the thermodynamics of a drug–polymer mixture: adding a third component alters the effective drug–polymer interaction, which in polymer science is called the "delta chi effect" (Pouchlý and Patterson 1976; Robard 1977). The paper asks which of the two explains water-induced AAPS: does water only free a mixture that was already thermodynamically unstable but kinetically trapped, or does it make an otherwise stable mixture unstable? C43

How each technique sees a phase. Four methods are combined.

• AFM topography maps the height of the film surface. Phase-separated films show pits or raised regions; a miscible film is flat. R2·R5·R6
• AFM-IR (nano-IR): a pulsed, tunable IR laser heats the sample when it hits an absorption band; the resulting thermal expansion makes the AFM cantilever ring, and the amplitude of this "ringdown" is proportional to absorption. Local spectra and chemical maps follow, with a lateral resolution of about 100 nm (Harrison 2013). The authors used 1590 cm⁻¹ (conjugated aromatic ring stretch of miconazole) as the drug marker and 1679 cm⁻¹ (carbonyl stretch of the pyrrolidone ring) as the PVPVA marker. P7·R3·R27·C8
• LCR: an alternating current through the cantilever in a magnetic field makes it oscillate. The contact resonance frequency depends on the stiffness under the tip, so driving the tip at the first flexural resonance of one material makes regions of that material bright. P8·C15
• CFM: a confocal microscope images one focal plane at a time; each dye is excited with its own laser line and read through its own emission filter, giving one colour channel per dye. P9

The resolution limit of the light microscope. The authors estimate the smallest discrete domain CFM can resolve from the Abbe diffraction limit (Section 4): E1

Here $d$ is the smallest resolvable feature, $\lambda$ is the wavelength of the light (taken as the prodan excitation wavelength, 407 nm) and $\mathrm{N}\mathrm{A}$ is the numerical aperture of the objective (1.49). In plain terms: the shorter the wavelength and the wider the cone of light the objective collects, the finer the detail it can separate. The authors give about 150 nm; with the printed inputs the expression gives about 137 nm (0.5 × 407/1.49), so the stated figure is a rounded estimate. E1·C37·X3 Either way, CFM misses domains smaller than roughly 140–150 nm, a coarser limit than nano-IR (about 100 nm) or the about 30 nm quoted for DSC; its advantage is speed, under 1 min per sample. R23·R27

Methods in detail

The study has one composition and many ways of looking at it. All films were 50:50 w/w miconazole:PVPVA (with pure amorphous miconazole and pure PVPVA films as references), and the experiments fall into three contributions: choosing a dye pair, checking the dyes against AFM-based chemistry and stiffness, and applying the dyes to three routes of water entry. DSC on a powder of the same composition adds glass transitions. A1·A2·A3·A4

Film preparation, the common starting point

Miconazole and PVPVA were dissolved together at 50 mg/mL solids in methanol:dichloromethane (50:50 v/v) and spin-coated in two stages (500 rpm for 6 s, then 3,000 rpm for 30 s). Two sets of solutions were made: one without dye for AFM work (15 µL on a ZnS substrate) and bulk IR (50 µL on KRS-5), and one with each dye at 0.01% w/v (100 µg/mL) for fluorescence work (100 µL on a 1 mm quartz slide). P1·M15 "Low RH" films were cast inside a glove box purged with dry nitrogen, at 20% RH; "high RH" films were cast in the laboratory, where the RH was recorded for each film and ranged from 45 to 60%. All films were dried overnight under vacuum and were estimated from AFM heights to be about 300 nm thick. P1·R1 The contrast between the two casting conditions is the paper's first experiment: in both cases solvent is present while the film forms, but only in the second is water from the air also present. A1·C43

Contribution 1: choosing a dye pair

Problem. The method needs one dye that stays in the drug-rich phase and one that stays in the polymer-rich phase, that do not fade under illumination, and whose excitation and emission bands do not overlap on the available microscope. P4

What they did. Eight dyes were added one at a time to humid-cast 50:50 films, which contain discrete and continuous phases, and their location was observed with a wide-field fluorescence microscope. Photobleaching was judged from three consecutive images with 500 ms exposure and no lag. The laser lines and emission filters of the confocal microscope (Nikon A1), the spectral distance between a drug dye and a polymer dye, and cost were also weighed. P4·R11 The full screening data are in SI Table S1, which was not available here; Table 1 of the paper summarises them (Table B below). R11

Why it should work. The idea follows classical fluorescent staining: a hydrophobic dye partitions into the hydrophobic (drug-rich) phase and a hydrophilic dye into the hydrophilic (polymer-rich) phase. A similar two-dye approach had been used to map polymer gradients in phase-separated polymer blends (Dibbern-Brunelli 1998; Granados 2001). C35·C7

Contribution 2: checking the dyes against chemistry and stiffness

Problem. A dye's colour shows where the dye is, not what the phase is made of. The phase identity has to be established by a method that sees the components themselves. C7

What they did. On the dye-free films, AFM topography was recorded in contact mode (5 × 5 µm fields for cast films; 256 × 256 points; 0.5 Hz). Nano-IR spectra were taken over 1200–1800 cm⁻¹ (256 co-averages, 4 cm⁻¹ spacing; each shown spectrum is the average of 3 and is Savitzky–Golay smoothed), and chemical maps were recorded with the laser tuned to 1590 or 1679 cm⁻¹. P5·P7 Bulk transmission FTIR of the pure films (128 scans, 4 cm⁻¹ resolution) served as reference spectra. P6·R4 For LCR, the cantilever was swept from 1 to 1000 kHz on the pure films to find each material's first flexural resonance, and the films were then mapped while driving the tip at the miconazole frequency (113.8 kHz) or the PVPVA frequency (112.6 kHz). P8·R9 The dyed films were imaged in the confocal microscope with a 60× oil objective (NA 1.49): prodan excited at 407 nm and read at 425–475 nm (shown blue), R6G excited at 561 nm and read at 570–620 nm (shown red), at 1024 × 1024 pixels. P9

Why it works, and its limit. Three independent signals (IR absorption at a component band, mechanical stiffness, and dye location) are compared on the same kind of film. Note by the AI writer (not discussed in the paper): they were not measured on the same film. The AFM methods used the dye-free set on ZnS and the confocal images the dyed set on quartz, so the dye assignment is checked by matching the patterns, not by co-localisation. P1·X14

Contribution 3: three routes for water

Films cast at 20% RH, and therefore miscible to begin with, were exposed to water in two further ways. A2 For storage, they were kept in a desiccator at 75% RH over saturated NaCl at room temperature and examined after 1, 2, 3, 7 and 28 days by CFM and AFM. P2 For immersion, they were placed in pH 6.8, 100 mM phosphate buffer for 5, 10, 20 and 30 min, rapidly dried and imaged by CFM; AFM topography and nano-IR imaging were done only for the 30 min sample, over 50 × 50 µm. P3·P5 Together with humid casting, this gives three routes in one system: water present while the film forms, water vapour slowly absorbed into the solid, and liquid water. A1·A2

Powders and DSC

A 50:50 powder was made by rotary evaporation from the same solvent (45 °C water bath, reduced pressure) and dried under vacuum overnight. About 10 mg portions were equilibrated at 75 or 100% RH in a vapour sorption analyser at 25 °C, until the weight changed by less than 0.01% in 5 min, or for at most 180 min. P10·P11 Glass transitions were measured by modulated DSC (3–5 mg in hermetically sealed pans, 2 °C/min from −40 to 120 °C, modulation 1 °C every 60 s), and reported as midpoints. One dry sample was first held at 100 °C for 60 min, cooled and rescanned. P12 The water content reached by the equilibrated powders is not reported. X8

Experimental analysis

Reference spectra and stiffness of the pure components

Spin-coated films of pure miconazole and pure PVPVA were smooth in AFM (Figure 2A). Their nano-IR spectra each carried a clear marker band, 1590 cm⁻¹ for miconazole and 1679 cm⁻¹ for PVPVA (Figure 2B), at the same positions as in bulk transmission FTIR (Figure 3); the authors describe the agreement as "good". R2·R3·R4·C8·C9 In LCR the first flexural resonance of PVPVA lay slightly below that of miconazole, 112.6 versus 113.8 kHz in the text (the plotted maxima in Figure 7B read about 113.6 and 112.3 kHz), which the authors read as PVPVA being the softer material. A 1.2 kHz separation was enough to give stiffness contrast in maps. R9·C16

Result 1: water in the air during casting decides one phase or two

The film cast at 20% RH was flat in AFM (height range ±20 nm over 5 × 5 µm) and gave the same nano-IR spectrum at five points across it, each showing both the drug and the polymer band (Figure 4). The authors take this as a miscible film at the roughly 100 nm resolution of nano-IR ("suggesting", "indicating"; the abstract says "confirmed"). R5·R27·C10

The film cast at 60% RH was different (Figure 5A). Its surface held discrete, circular pits about 200–400 nm deep in a continuous phase; in the 5 µm field the pits are about 0.3–0.7 µm across, with some merged into larger regions. R6 Spectra from a pit and from the matrix both contained both marker bands, which the authors say reflects partial miscibility and/or spectral mixing, either through the depth of the film or between domains closer than the resolution. R7·C11 Because local IR intensity depends on the height of the feature, they compared relative peak heights of normalised spectra: the 1590 cm⁻¹ band was relatively higher in the pits and the 1679 cm⁻¹ band in the matrix (Figure 5B), so "the discrete phase is miconazole-rich, whereas the continuous phase is PVPVA-rich". R7·C12·C13 Chemical maps at the two wavenumbers made the same point over the whole field: the pits were bright at 1590 cm⁻¹ and dark at 1679 cm⁻¹ (Figure 6). The text of Section 3.1.2 cites these maps as "Fig. 8", but they are Figure 6; Figure 6 is never cited by its own number. R8·C14·X1

LCR gave a third view, on a film cast at 45% RH (Figure 8). Its pits were smaller, about 0.2–0.5 µm, and shallower (height range ±40 nm). Driving at the miconazole resonance (113.8 kHz) made the discrete spots light; driving at the PVPVA resonance (112.6 kHz) made them dark and the matrix light. The authors say this "further support[s] the supposition" that the discrete domains are drug-rich. R10·C17·X13

Finally, the dyes. In the 20% RH film both dyes spread evenly: uniform blue at 407 nm, uniform red at 561 nm, and purple when combined (Figure 9); the caption states that the homogeneous fluorescence "confirms" miscibility. R12·C21 In the 60% RH film the dyes separated sharply: approximately circular blue (prodan) domains in a red (R6G) continuous phase (Figure 10). R13·C22 The discussion gives the domain size after casting at moderate RH as "submicron to few microns"; in Figure 10 the largest blue domains measure about 15 µm against the 20 µm scale bar. R13·C39·X5

Table A. What each film looks like under each method. Sizes were not measured by the authors; they are readings from the scale bars and axes of the figures (approximate). R5·R6·R7·R8·R10·R12·R13·R15·R16·R18·R19·R20

Result 2: hydrophobic dyes go to the drug, charged dyes to the polymer

Table B. Dye screen in humid-cast 50:50 films (from Table 1 and Sections 3.1.4.1 and 4). Laser line and filter are those available on the Nikon A1 confocal microscope. R11·M3·M4·M5·M6·M7·M8·M9·M10

The relatively hydrophobic dyes (prodan, pyrene, nile red) went to the discrete, drug-rich domains, and the two charged dyes (cationic R6G and anionic Alexa Fluor 488) to the continuous, polymer-rich phase, as the staining principle predicts. R11·C18·C35 The exception was the fluorescein family: fluorescein, FITC and 5-DTAF, relatively hydrophilic because of their ionizable carboxylate, also went to the drug-rich phase, which the authors call "somewhat surprising". They suggest that this preference "may arise" from specific interactions between acidic fluorescein and weakly basic miconazole; no measurement addresses it. R11·C18·C36 Pyrene, nile red and the fluorescein dyes photobleached, and pyrene (excitation maximum 350 nm, emission 380–400 nm) had no matching laser line. A usable pair must also have separate excitation and emission bands, since overlap between channels "could lead to imaging artifacts"; with cost included, prodan and R6G were chosen. R11·M5·C19·C20

Result 3: a week at 75% RH splits an initially miscible film

Films cast at 20% RH and stored at 75% RH showed phase separation after 7 days, in confocal images and in AFM-IR (Figures 11 and 12). As after humid casting, the drug-rich phase formed discrete domains, about 1–3 µm in the confocal images and pits of about 1.5–3 µm in AFM, in a polymer-rich continuous phase; in AFM-IR the pits were bright at 1590 cm⁻¹ and dark at 1679 cm⁻¹. No further change in morphology was seen up to 4 weeks. R15·R16·C24 The methods list evaluations after 1, 2 and 3 days as well, but the results report only the 7-day state, so when the separation began is not known. P2·X7

Result 4: in buffer, a coarse and connected structure within 30 minutes

Immersion gave a different picture (Figure 13). After 5 min the R6G channel showed bright, discrete domains of about 4–8 µm in a prodan-stained matrix: the polymer at first "appears" to form the discrete phase in a drug-rich matrix. R18·C28 At 20 min, large prodan-stained regions tens of micrometres across, each holding a few small R6G-bright inclusions, were separated by R6G-stained continuous regions. At 30 min, prodan-stained domains of about 10–25 µm were separated by an interconnected R6G network, which the authors describe as "more bi-continuous". R18·C28 AFM at 30 min confirmed a bicontinuous topography over 50 × 50 µm, with higher and lower regions about 5–15 µm wide (Figure 14), and AFM-IR showed that the higher regions absorbed more at 1590 cm⁻¹ and less at 1679 cm⁻¹, so they are drug-rich (Figure 15). R19·R20·C27

The authors explain the height difference as "most likely" the result of the polymer-rich phase dissolving faster than the drug-rich phase once the phases have formed. C29 They also note that the length scale of separation becomes large: immersed films show a much coarser microstructure than films exposed to high RH, and the phases "undergo considerable rearrangement" with longer immersion. C45·C46 Two points of reading: the caption of Figure 13 repeats the wording of the earlier captions ("Discrete phase is drug-rich as shown in (B)"), which matches neither the images nor the text, so the figure is best read by channel, blue for drug-rich and red for polymer-rich; and the films were dried before imaging, so the panels are dried end points rather than the structure in buffer. X2·X12

Result 5: one glass transition dry, two after humidity, one after heating

Table C. Modulated DSC of the 50:50 rotary-evaporated powder (Sections 3.1.5, 3.2, 3.3 and 4.1); humid samples equilibrated at 25 °C. M1·M2·R14·R17·R21·R22·C23·C25·C30·C44

The dry powder had a single Tg of 26 °C, between those of miconazole (0 °C) and PVPVA (107 °C). R14·M1·M2·C23 After equilibration at 75% RH it showed two Tg events, 10 and 48 °C, and after 100% RH two at −5 and 68 °C; the authors assign the lower one to the drug-rich and the upper one to the polymer-rich phase. R17·R21·C25·C30 When the dry powder was held at 100 °C for 60 min, about 75 °C above its Tg, and rescanned, it still had a single Tg; the trace is reported as "data not shown". R22 Note by the AI writer (not discussed in the paper): the drug-rich Tg at 75% RH (10 °C) lies above that of pure miconazole, so that phase still holds polymer; the polymer-rich Tg is higher after 100% RH (68 °C) than after 75% RH (48 °C), the opposite of what extra water alone would do; and without water contents for the powders neither trend can be interpreted. X8

Discussion

Is water a source of mobility or a change in thermodynamics?

The authors' central argument (Section 4.1) concerns the role of water. One view is that water only raises molecular mobility in a mixture that was already thermodynamically inclined to separate but kinetically trapped. The other is that absorbed water changes the composition, and with it the thermodynamics, so that a stable binary mixture becomes unstable (the "delta chi effect"). C43 Two observations lead them to the second view. First, both the dry-cast and the humid-cast films were formed from solution, where mobility "is likely" high because solvent is present, yet only the humid-cast film separated; "evidence is provided for the thermodynamic role of water". C43·R5·R6 Second, the dry powder stayed one phase after 60 min at 100 °C, about 75 °C above its Tg; "This suggests that mobility is not the sole criterion required for phase separation". C44·R22 They add that this "conjecture" fits two observations from the literature: miconazole–PVPVA dispersions made by melt extrusion are miscible, and phase-separated ones can remix under compression (Singh 2014, 2015). C44·R24 The conclusion is that miconazole and PVPVA "do not appear to be inherently immiscible"; the separation "is induced by the presence of water, which is an antisolvent for the hydrophobic miconazole". C41·C42 The same reading is applied to immersion, where the separation is "also consistent with water changing the thermodynamic properties of the system". C45

The flow below lays out the chain as the authors give it.

Note by the AI writer (not discussed in the paper): neither branch is marked better or worse here, because the paper measured no release or stability outcome for either structure.

Why the routes give different structures

The authors expect that a system which separates when water enters by one route "will likely" also separate when it enters by another, but "given the different thermodynamics and mobilities, the evolution of the microstructure may differ considerably", and they call for a systematic comparison across routes. C34 This paper supplies such a comparison for one system. Humid casting and humid storage both gave discrete drug-rich droplets in a polymer-rich matrix, which the authors suggest "may occur via a nucleation and growth mechanism" (Sagui and Grant 1999). C24·C40 Immersion gave polymer-rich droplets in a drug-rich matrix at 5 min and then a coarse bicontinuous network, on a length scale an order of magnitude larger than after storage (Table A). R15·R18·C46 They do not offer a mechanism for this switch beyond faster water uptake and faster dissolution of the polymer-rich phase. C29·C45

The method itself

The authors place two-dye CFM as "a useful addition to the toolbox of analytical methods for miscibility assessment": it is fast (under 1 min per sample), sample preparation and interpretation are simple, and the theoretical resolution limit is about 150 nm. C37·R23·E1 They note that the same approach has already flagged phase separation in itraconazole–HPMC dispersions, checked by solid-state NMR (Purohit 2017), as "evidence of versatility". C38·R25 In the abstract it is "a promising new approach", and in the conclusions it "will ultimately be useful to improve understanding and optimize the formulation" of ASDs. C49 They also suggest that the phase behaviour of initially miscible ASDs during dissolution "may well be of great importance" for drug release, so methods like this "are likely to be of great importance" when comparing formulations. C47

What it suggests for formulation

Stated by the authors: if water is kept out while a solvent-based ASD forms (low RH), a miscible film can be obtained ("It is clear"), whereas moderate RH during film formation gives drug-rich domains; they offer this as "perhaps" some insight into what governs the miscibility of solvent-made ASDs, including the spray-dried miconazole–PVPVA dispersions that separated above 20% drug. C39·R24

Note by the AI writer (not discussed in the paper): three practical points follow for a formulator. Ambient humidity during solvent casting or spray drying is a process variable for ASDs of hydrophobic drugs: room air at 45–60% RH was enough to separate a 50:50 film that was miscible at 20% RH. P1·R6·R5 A single Tg on a dry powder does not show that the dispersion stays one phase in use: the same powder had two Tgs after 75% RH. R14·R17 And the structure formed in buffer, with connected drug-rich regions left behind as the polymer leaves, is the kind of structure that could slow drug release from high-loading ASDs, although release was not measured in this paper. R18·R20·C47

Limitations

Limitations stated by the authors. The authors are explicit about the resolution of CFM (about 150 nm by their estimate), so domains finer than that are missed. C37 They note that both marker bands appear in both phases of the humid-cast film, which may reflect partial miscibility or spectral mixing, and they compare normalised relative peak heights for that reason. C11·C12 They describe the immersion morphologies with hedges ("it appears", "appear more bi-continuous") and the dissolution explanation with "most likely". C28·C29 The thermodynamic role of water is argued from the available observations and called a "conjecture". C43·C44

The AI writer's own judgement (1): what the evidence covers. The qualitative result is well supported: without water the 50:50 films are one phase, with water they are two, and the drug-rich phase is identified by nano-IR spectra, nano-IR maps and LCR maps, which agree with each other. R5·R7·R8·R10 Three points weaken the rest. The dye assignment rests on matching patterns between separately cast films, not on measuring the same film by both methods. X14 The thermodynamic interpretation rests on an argument about mobility during casting, which was not measured, and on an annealing DSC run that is not shown; no water content, phase composition or interaction parameter was determined. X9 And the claim in the conclusions that water during storage led to "rapid" phase separation is not shown, because only the 7-day result is reported. C48·X7

The AI writer's own judgement (2): reading the images. The images are qualitative, and the films differ between figures. The "high RH" films were cast at 45–60% RH, and the LCR film (45% RH) has smaller and shallower pits than the nano-IR film (60% RH). X13 Domain sizes were not measured; the discussion's "submicron to few microns" does not cover the domains of up to about 15 µm in Figure 10, and the scale bars imply confocal fields of about 47–93 µm wide, whereas the stated 0.07 µm per pixel over 1024 pixels gives about 72 µm. X5·X6 The film thickness (about 300 nm) is smaller than the deepest pits (up to 400 nm) without the paper saying which film was measured. X4 Immersed films were dried by an unstated method before imaging, the 10 min sample is not shown, and the Figure 13 caption contradicts the text. X12·X2 The text of Section 3.1.2 refers to the AFM-IR maps as Figure 8, which is the LCR figure. X1 Finally, all of this concerns one drug, one polymer and one composition (50:50), and the DSC was done on powders rather than on the films that were imaged. X8

Conclusion

The paper shows that a pair of dyes with opposite preferences, prodan for the drug-rich phase and rhodamine 6G for the polymer-rich phase, turns a confocal microscope into a quick map of drug–polymer phase separation in miconazole–PVPVA films, and that the map agrees with AFM-based IR and stiffness mapping on parallel films. R11·R13·R8·R10·C48 With it, the authors follow water entering the same 50:50 dispersion in three ways: humid air during casting and 75% RH storage give discrete drug-rich droplets in a polymer-rich matrix, while buffer gives, within 30 min, a coarse bicontinuous structure with raised drug-rich regions. R13·R15·R18·R20 Dry films and powders, even after heating 75 °C above Tg, stay one phase, which the authors read as water acting through the thermodynamics of the mixture rather than through mobility alone: "the phase separation is induced by the presence of water". R5·R14·R22·C42·C44 What remains open is how much water is needed, how fast separation starts on storage, and whether the structure formed in buffer changes drug release. X7·X9·C47